Using non-medical risk factors related to dementia and cognitive decline for developing an evidencebased e-health tool

(1)

IN

DEGREE PROJECT INFORMATION AND COMMUNICATION TECHNOLOGY,

SECOND CYCLE, 30 CREDITS STOCKHOLM SWEDEN 2016,

Using non-medical risk factors related to dementia and cognitive decline for developing an

evidencebased e-health tool

ANUSHARANI GOPU

KTH ROYAL INSTITUTE OF TECHNOLOGY

(2)

TRITA TRITA-ICT-EX-2016:186

(3)

Using non-medical risk factors related to dementia and cognitive decline for developing an evidence-based

e-health tool

By

Anusharani Gopu

Master thesis submitted in partial fulfilment of the requirement for the degree of Master of Science in Distributed Systems.

(School of ICT, KTH)

Supervisor: Marie Sjölinder

Examiner: Magnus Boman

(4)

(5)

Acknowledgement

First and foremost, I would like to thank my examiner Prof Magnus Boman, without whom it wouldn't be possible to complete this thesis work. He was kind and patiently answered all my queries without any delays and was accessible throughout the summer when I did most of the work. I also would like to thank my supervisor Marie Sjölinder for her input and valuable suggestions mainly regarding the structure, integrity and quality of the report. I am also thankful to Caroline Wärn, Mattias Jacobsson and all other people at SICS for their valuable input and support.

Thanks to Shireen Sindi and Rui Wang at Aging Research Centre (ARC), Karolinska Institute, for their valuable time in explaining risk score calculations and statistical modelling. Special thanks to my employer Bo Öström at Bogleboo AB who granted me leave from my duties to perform this thesis work. Last but not least, I would like to thank my family for giving me the moral support during the thesis work.

Anusharani Gopu KTH, September 2016

(6)

Abstract

The number of dementia cases is increasing worldwide. Most research and development in this area is related to the prevention of dementia, and to the development of various prediction tools for dementia. The tools made available take most of the medical data into account while calculating risk scores, with only a small amount of non-medical data. There is a lot of data related to medical and non-medical risk factors available from various sources which can be retrieved and analysed in real time, but this is today not used in any risk score tool for risk score calculation. As part of the project Multimodal strategies to promote a healthy brain in ageing:

Innovative evidence-based tools (MULTI-MODE), a new risk score is being developed to be used in a new ICT-based tool for dementia prediction. Identification of non-medical data and a good model to fill the gap between data available at the server and using this data in risk score calculation may help in increasing the predictability of tools. In this thesis, some of the existing risk factors for the prediction of dementia are described, and the importance of non-medical factors in calculating risk scores is discussed. Additional non-medical factors are identified that could be included in future versions of the risk score. A database design for storing risk score information efficiently is presented, as is an app structure that can be used at the server side to validate the user input and to increase the effectiveness of a prediction tool.

Keywords

Alzheimer’s disease, Dementia, Mild cognitive impairment, Risk factors, Dementia risk score, Cognitive decline, Cognitive training, Dementia prevention, Evidence-based prediction, Risk score app.

(7)

Sammanfattning

Antal demensfall ökar över hela världen. Forskning och utveckling inom detta område är relaterat till att förebygga demens och att utveckla olika prognosverktyg för demens. Flera tillgängliga verktyg tar hänsyn till medicinska data i beräkning av riskpoäng, med endast en liten mängd av icke-medicinska data. Det finns en hel del data om medicinska och icke-medicinska faktorer online, men de används idag inte för riskpoängberäkning. Som en del av projektet Multimodala strategier för att främja en frisk hjärna i åldrande: Innovativa evidensbaserade verktyg (MULTI-MODE), så har en ny metod utvecklats för att användas i ett nytt IT-baserat verktyg för demensförutsägelse. Identifiering av icke-medicinska data och en bra modell för att överbrygga gapet mellan tillgängliga data på servern och använda dessa data i riskberäkning kan bidra till att öka precisionen hos verktyg. I den här studien beskrivs en del befintliga riskfaktorer för förutsägelse av demens och vikten av icke-medicinska faktorer i beräkning av risk diskuteras. Ytterligare icke-medicinska faktorer identifieras som skulle kunna ingå i framtida versioner av riskverktyg (såsom appar). Vissa identifierade riskfaktorer har analyserats och visade att effekten av att införa icke-medicinska faktorer ökar precisionen i resultaten. En databasdesign för lagring av riskinformation på ett effektivt sätt presenteras, liksom en appstruktur som kan användas på serversidan för att validera några av de parametrar som kan öka effektiviteten av verktyget.

Nyckelord

Alzheimers sjukdom, Demens, Kognitiv svikt, Riskfaktorer, Kognitiv träning, Demensförebyggande, Evidensbaserad förutsägelse.

(8)

List of figures

Figure 1 Venn diagram representing AD as subset of Dementia, Dementia as subset of Mild cognitive

impairment (MCI). ... 4

Figure 2 Classification of risk factors related to dementia. ... 7

Figure 3 Cost per Genome. ... 10

Figure 4 Percentage of patient population where a particular drug in a class is ineffective. ... 11

Figure 5 CAIDE Risk calculator Personal Data. ... 16

Figure 6 CAIDE Risk score calculator level indicator graphs. ... 16

Figure 7 CAIDE Risk score calculator information pages. ... 17

Figure 8 Algorithm for the old CAIDE risk score. ... 28

Figure 9 Suggested Object/Class diagram for risk score app development including the relationships between different objects. ... 33

Figure 10 Database relationship diagram for storing user’s basic information at server. ... 34

Figure 11 Database relationship diagram for storing caregiver information at server. ... 35

Figure 12 Database relationship diagram for storing User’s device information at server. ... 35

Figure 13 Database relationship diagram for storing user’s risk factors with different timestamps at server. ... 36

Figure 14 Database relationship diagram showing relations between user, caregiver and users calculated risk score with different timestamps. ... 36

Figure 15 Sequence diagram: Retrieval of DBP from Caregivers Database for Risk score calculation. ... 37

(11)

List of tables

Table 1 CAIDE Risk Score ... 14

Table 2 Comparison of risk scores ... 18

Table 3 LOAD Risk score ... 20

Table 4 Late-life dementia risk score ... 21

Table 5 Example data entered by a fictive person into Risk App ... 29

Table 6 Risk scores for fictive data in table 5 ... 29

Table 7 Example data for Air Quality Index ... 30

Table 8 Example Javascript function which demonstrates the calculation of user’s age using Swedish personal number or date of birth of the user ... 32

(12)

List of abbreviations

Abbreviation Explanation

AD Alzheimer’s disease

ICT Information and communications technology

KI/ARC Karolinska Institutet/Aging Research Centre

CAIDE Cardiovascular Risk Factors, Ageing and

Dementia

SNAC-K Swedish National study on Aging and Care in

Kungsholmen

SICS Swedish Institute of Computer Science

RISE Research Institutes of Sweden

MULTI-MODE Multimodal strategies to promote a healthy

brain in ageing: Innovative evidence-based tools

EIT European Institute of Innovation and

Technology

APOE Apolipoprotein E

BMI Body Mass Index

FINGER Finnish Geriatric Intervention Study

NTB Neuropsychological Test Battery

ANU-ADRI Australian National University AD Risk Index

SBP Systolic Blood Pressure

DBP Diastolic Blood Pressure

IoT Internet of Things

RCT Randomised control trial

AQI Air Quality Index

My-AHA My Active Healthy Ageing

SNP Single-Nucleotide Polymorphism

(13)

(14)

1 Introduction

This study is of risk factors related to dementia, particularly on non-medical risk factors that can be considered while developing evidence-based tools for predicting dementia risk. A risk factor is something that increases the likelihood of developing a condition or disease.

Dementia is caused by neurodegeneration and is characterised by a disturbance of multiple brain functions, including thinking, memory, calculation, orientation; language, learning capacity, and judgement which eventually make it difficult for people to perform daily activities [1]. Alzheimer’s is the most common type of dementia and possibly contributes to 60-70% of cases [2]. Other types of dementia are vascular dementia (e.g., strokes, heart failure), dementia with Lewy bodies (e.g., sleep behaviour disorder), and frontotemporal dementia (e.g., language disturbances, deterioration in behaviour and personality) [3].

Brain-related health issues at late-life will burden societies and family members physically, psychologically, and economically. There are around 34 million individuals worldwide suffering from Alzheimer’s disease (AD) and there is a prediction that this number will be tripled in the coming 40 years, due to changes related to demography and an increase in life expectancy [4].

Along with memory loss, Alzheimer’s patients can experience other symptoms like difficulty in judgement, thinking, and confusion related to place and time. In some cases, patients cannot carry out basic functions like swallowing and walking. This increase in dementia at older age requires more attention towards elderly people from formal caregivers (community care professionals) or from informal caregivers (friends or relatives), which also increases society health care costs tremendously [5] that contributes to one-third of all health related expenses.

Costs related to dementia and AD in Europe were estimated to > €170 billion in 2008 [6].

Governments and caregiving institutions have become aware of the impact of brain problems in at-risk elderly people towards family members, societies, and individuals [7], prompting more research into decreasing future health costs through prevention. Such prevention is possible with diagnosing and treating these diseases at early stages. Information and communications technology (ICT) can play a very important role in the prediction of these ageing problems by developing applications (related to measuring diet, exercise, weight loss and fitness), tools, and sensors which can be used for early detection of these diseases [8].

Dementia impact has received the greater attention of governments and politicians all over the world in recent years. Governments of developed countries like UK, Norway, Sweden, France, U.S., and South Korea have developed specific plans or strategies [9]. Dementia cases are usually concerned with elderly people but there is an increase in early-onset dementia (under 65 years of age). Although there is an increase in early-onset dementia cases, dementia is a condition which affects elderly people and is the leading contributor to disability and dependence [10], [11]. Rapid increases in the numbers of older people are forecast for China, India, and Latin America. Dementia awareness and a health system that can tackle the dementia related problems are much more limited in less developed regions [12]. It is therefore very important to track the global pervasiveness of this difficult situation and to provide findings related to dementia. As there are no curative treatments for dementia, prevention has been highlighted as the major health priority according to the G8 dementia summit and World Health Organization [13], [14]. Intervention towards these modifying risk factors helps in delaying the onset of AD [15], and it has been estimated that 10-25% reduction in key risk factors could prevent 1.1-3.0 million AD cases internationally [15]. Therefore, it is helpful to have evidence-

(15)

based tools that measure individual’s risk factor more accurately and thereby start the treatment earliest in order to prevent the dementia progress. Prediction of risk analysis at early stages is helpful for both patients and their caregivers as they get to benefit from counselling on how to handle the diseases, and so can they plan. Early diagnosis is also helpful for proper medication.

To investigate the effect of risk factors on problems that occur in late life like dementia and cognition, the Department of Neurology, University of Finland, the National Institute of Health and welfare, Helsinki, and KI/ARC (Karolinska Institutet/Aging Research Centre) together started a research called CAIDE (Cardiovascular Risk Factors, Ageing and Dementia) and identified a number of factors that may increase or decrease the risk level of dementia or cognitive impairment. These risk factors were used to develop CAIDE risk score. The CAIDE Dementia Risk Score tool has been developed to predict the chance of occurrence of dementia in a person in next 20 years [16]. This tool was developed based on considering various risk factors like age, hypertension, physical activity, obesity, and educational background. The CAIDE dementia Risk Score tool has been validated in a multi-ethnic population in the U.S. In a study made by ANU-ADRI, the CAIDE risk score was calculated and compared in three different older age cohorts, and validation of result was found to be poor [17]. This was likely due to CAIDE development having been done by considering mid-life risk factors and validated in older cohorts. This, in turn, suggests that risk factor effects can be different in different cohorts. The CAIDE risk score was developed by considering midlife factors which may not be suitable for older cohorts. There is thus a need for a new risk score for predicting dementia in older cohorts.

Researchers at KI/ARC have developed an algorithm for a new risk score that can be applicable for the older population, based on data collected from Swedish National study on Aging and Care in Kungsholmen [SNAC-K] project [18]. As a part of this project, researchers at Swedish Institute of Computer Science (SICS) are going to produce evidence-based ICT tools to predict dementia risk and prevent cognitive decline/dementia, based on the new risk score. These tools are intended to be used by citizens as well as by healthcare staff. The main purpose of developing these tools is to reach more users, reducing health costs, social burden, and to provide better tools for dementia prediction.

1.1 Problem

Developed risk scores for dementia take mostly medical data (ex. blood pressure, cholesterol), and only a limited amount of non-medical data (ex. age, gender, diet, smoking), into account when calculating individual risk, whereas excluded non-medical data may in some cases play a significant role, and could arguably contribute to the risk score.

1.2 Purpose

There is an opportunity for ICT-based tools to take non-medical data into account that may help in developing better predictive ICT tools for dementia. The purpose is to identify such non- medical data that can be included in future risk score calculation, and a suggested database design for storing risk score information. The study can be considered a complement to the MULTI-MODE (Multimodal strategies to promote a healthy brain in ageing: Innovative evidence-based tools) project agenda, in which researchers will develop and validate innovative ICT tools, notably an app, to help predict risk scores for dementia.

(16)

1.3 Goal

The goal is to contribute to an increase in the usefulness of ICT tools that predict cognitive decline, in part through the identification of non-medical data that can be included in risk score calculation, enhancing risk score by checking the user’s input accuracy, and in part through a suggested database design for storing risk score information efficiently.

1.4 Methodology

The SICS, as one of the partners in MULTI-MODE, and a part of Research Institute of Sweden (RISE), was assigned to develop and test eHealth tools to promote healthy brain ageing and widely implement the risk scores. The thesis work was carried out at SICS. Interaction with SICS researchers part of MULTI-MODE was instrumental to this work, including participation in internal meetings and an internal workshop in May 2016. Some advice and communication with KI/ARC researchers were also helpful to obtaining the results.

As part of MULTI-MODE, tools will be designed for use by citizens and health care staff, making use of state-of-the-art mobile technology. The tools will be based on the new risk score derived from CAIDE Dementia Risk and will be validated in different datasets. The datasets are from many longitudinal cohort studies like SNAC, KI, and CAIDE, Rotterdam study, United Kingdom Clinical Practice Research Datalink and CHARIOT register. 8000 Swedish citizens participated in SNAC study, 2000 participated in CIADE, and CHARIOT register involved 26,000 individuals.

Apart from this data, risk score also considered the data from medical records, quality registers (ex. workshops, reviews, inspections, and audits), and prescribed drug registers. Developed new risk score has also used the data from H2020 Athlos project which has data related to more than three hundred thousand individuals.

At the beginning stage of the thesis work, a basic understanding of the problem background and the former research was achieved through the documents provided by Professor Magnus Boman at SICS who is the SICS project leader for MULTI-MODE. As the field of dementia risk assessment is relatively mature, a deductive literature study and desktop research were completed by collecting information regarding risk scores and other factors related to problems that occur in late life. Interviews have been conducted with KI/ARC researchers to establish the process of developing the first CAIDE risk score, which was based on mid-life risk factors. These interviews helped in understanding in depth the existing risk score but still not able to find the algorithm used in OLD CAIDE risk score. So an attempt made to produce an algorithm used for old CAIDE risk score in Chapter 6.1. For knowing about a newly developed risk score, a semi- structured interview has been conducted with Rui Wang, PhD at ARC. After studying and understanding existing risk scores it has been observed that the majority of the risk factors involved in the risk score calculation are medical-related risk factors. There is relatively little importance given to non-medical risk factors. Therefore, further study of non-medical factors was motivated. A reductionist approach has been used for identifying some additional risk factors, in which risk factors are categorised into two major categories which are further divided into sub-categories. The scope of the thesis is limited to the discussion on some of the non- medical factors instead of all identified factors and some assumptions were made about the risk factors data during the analysis. A qualitative approach has been used for understanding and identifying underlying non-medical risk factors.

(17)

Figure 1 Venn diagram representing AD as subset of Dementia, Dementia as subset of Mild cognitive impairment (MCI).

In this thesis, AD refers to studies associated only with Alzheimer’s, if a study relates to serious cognitive decline in general then it is considered as dementia, and other studies related to cognitive impairments beyond those expected based on the age and education levels, but which does not affect daily life are considered as mild cognitive impairment (MCI) [19]. MCI can be seen as an early symptom of AD but it is not dementia as it does not considerably affect daily life [20].

1.5 Thesis overview

Chapter 2 describes the role of ICT in helping elderly people, an increase in usage of mobile apps in the health sector, and various risk factor categories for dementia. Chapter 3 describes the new technologies developed within ICT and medical informatics, in the context of genomics, i.e. genetic risk factors. The chapter includes the benefits of personal genomics, precision medicine, and also provides a discussion about current trends in personal genomics. This chapter also describes the possibilities of introducing genome data in risk score applications.

Chapter 4 includes a discussion of different risk scores and analyses of these risk scores. Here, the CAIDE risk score is experimented with on some specific user data to understand how the existing score appears for users. The chapter describes the MULTI-MODE project that is currently undertaken within EIT (European Institute of Innovation and Technology) Health.

Chapter 5 describes the additional non-medical factors which may help in increasing the effectiveness of risk score. The effect of considering some of the identified non-medical factors into the risk score, a suggested database design for storing risk score information, and an app architecture is discussed in chapter 6. Chapter 7 contains a discussion on app-related issues, followed by conclusions.

(18)

2 The role of ICT in active ageing and risk factor categorisation

Dementia is a condition that affects mostly elderly people. It may affect younger people and dementia in younger people is called as early-onset dementia, but early-onset dementia cases are few when compared to late-onset dementia. Ageing of the population is growing very fast in both developed and developing countries. Europe is one of the major regions of the world where ageing of the population is advanced [21]. Population ageing occurs when the median age rises and shifts the distribution of a country’s population towards elderly people. The median age of the population is high in Europe when compared to other parts of the world.

Approximately 12.3 percent of the global population is people aged over 60 years and this may increase to 22 percent by 2050 [21]. This is going to have a dramatic effect on all aspects of society, especially healthcare. This increase in the ageing population may be a result of a decrease in fertility rate and increase in life expectancy. Monetary costs related to dementia represent a great financial burden to society. Furthermore, care towards the dementia patients is frequently associated with physical, social, and psychological stress [22]. So prevention is the best way to reduce these care costs to a greater extent, and as there are no curative treatments for dementia, prevention has been highlighted as the major health priority [13], [14]. ICT helps in prevention and better treatments for diseases. Following section explains how collaborating ICT with health helps in better elderly one’s life.

2.1 The role of ICT in active ageing

“ICT can help older individuals to improve the quality of life” according to “EU action plan on ICT and Ageing” [23 page 50]. Examples of ICT include using an app for improving diet which logs and monitors intake so that users can get regular advice about their dietary condition. One can make use of ICT to support cognitive deficiencies using sensors and actuators which can be helpful for elderly people by providing alternatives to their physical actions, such as moving or changing the position of objects (from closing doors to turning lights on and off), motion sensors, which helps in tracking a demented person. The concept of Internet of Things (IoT) (connected smart devices which communicate with each other and with software running on the cloud) also helps in taking actions, such as controlling heating and air conditioning, locking doors and windows, and giving reminders about medication. IoT technology securely collects and analyses the data from sensors and other devices. This information can be accessed by family members, healthcare professionals, or emergency services. Emergency services use the collected information, to help in emergency situations like falling of room temperature below a safe threshold which can be a life-threatening situation. This technology is very helpful for family members who live far away from elderly relatives. The policy makers of current health care systems are not only looking for efficient care solutions, but also for changing its focus from care to prevention to reduce care costs, burden on caregivers/family members, and burden on society. As part of this, current health care system focusing on preventing dementia with help of ICT by developing e-health tools for prediction of the brain related risk factors in early stages so that making these tools widely used by citizens and healthcare staff. EIT-Health and My-AHA (My Active Healthy Ageing) are the initiatives of EU which are both part of the Horizon 2020 effort, to provide good health and healthy active ageing in adults. In My-AHA, existing platforms (i.e., sensors to detect physical actives and apps to store this data) which already are owned by My-AHA partners will be integrated for the collection of data in large- scale and analyse the collected data using existing software or by improved firmware and

(19)

software. The type of data includes data related to sleep patterns, physical activities, diet, social activities, emotions, and clinical data such as blood pressure, sugar levels, Body Mass Index (BMI) etc. The data analysis results will be used by My-AHA for improving the activities and thereby reduce the “frailty risk” in older adults [24]. Apart from this Gerontechnology is one of the ICT technology which is devoted to the design of technology and environments for a better quality of life of older adults [8]. More information on Gerontechnology can be found in Appendix A.

2.2 The role of mobile apps in the health sector

In last decade the mobile app usage has increased tremendously. Mobile devices are easy to carry and are available at user’s fingertips whenever and wherever they need them. According to a recent report, in 2015 there were 120 000 mHealth apps (mHealth is the term used for the practice of medicine and public health supported by medical devices) that were available for download on AppStore. Out of 100 most downloaded apps from all categories, 33% apps were mHealth related. The report also mentions that most of these apps are very successful in terms of user satisfaction. mHealth apps are very popular right now especially in Sweden and they contribute to around 5.1% of total app downloads [25].

In order to succeed with an app, the app should meet the primary aim for which it is designed and developed for; all other extra features are nice to have in but should not affect the primary purpose. An example is ”Whatsapp” which is a simple texting app where users can be in touch with their family and friends. Despite many other advanced messengers, video chats and social media sites, this app is a huge success as it has very simple and easy GUI which meets the primary purpose i.e. sending out text messages very reliably and efficiently.

In the same manner, if designed properly, mHealth apps can be very useful for the elderly population by introducing them to physical and social activities and guiding them to overcome dementia-related problems. The apps can be helpful even for the caregivers to follow their patients in an efficient way and thereby increase their lifespan. A new mobile app is being developed as part of MULTI-MODE project that shows the dementia risk prediction for the elderly people.

2.3 Role of risk factors in risk calculation

Some people have a higher risk of developing dementia while others have a lower risk of developing dementia. Dementia risk assessment tools are helpful for the users for estimating the risk of being demented in future. The risk assessment tools use risk score calculation that is developed based on risk factor data from various studies. Understanding of risk factors related to disease may helpful in the prevention of disease. A risk factor is anything that increases the risk of a person developing a condition. Especially for dementia, there are a lot of risk factors – some of them are under person’s control (for example smoking, exercise etc.) and some of them are not (for example genes of a person, family history, age etc.). One cannot say that persons having any of the risk factors will necessarily develop dementia, in the same way avoiding any of the factors may not help the person stay healthy, but there are more chances of being healthy by avoiding certain risk factors. As discussed in chapter 1.4, this study chooses a reductionist approach as risk factors related to dementia can be categorised into three different categories. 1) Genetic risk factors 2) Medical factors 3) Non-medical risk factors.

Category three is further divided into two categories.

(20)

Figure 2 Classification of risk factors related to dementia.

Genetic risk factors related to dementia may include the APOE*E4 which has been associated with increase in risk of late-onset Alzheimer’s, and the three genes Amyloid Precursor Protein (APP), Presenilin-1 (PS-1) and Presenilin-2 (PS-2) which are associated with early onset Alzheimer’s, medical risk factors may include blood pressure, body mass index (BMI), cholesterol, diabetes, hypertension, etc. Non-medical risk factors are the ones which are external to the person’s body, only indirectly affecting health. This may include age, sex, educational years, alcohol consumption, smoking habit, etc.

Genetics is an important dementia risk factor which is not under person’s control. How the information related to genes is helping in treatments and prevention of dementia and specific genes by including in ICT tools is discussed in the following chapter (Chapter 3). There are a lot of medical, non-medical factors that effects dementia, but finding all those factors is out of this project scope. Instead, some of the non-medical factors have been identified (Chapter 5) and discussion of these factors is presented.

(21)

3 Targeting the individual

Most of the mobile apps like diet-related apps, risk prediction apps, etc. are targeted towards an individual, so it is important to see that individual as a part of wider development. Tools developed earlier did not give much importance to data related to individuals as most apps were developed from statistical and population-based studies. New technologies developed by ICT helps in improving the quality of life and reducing the effects of ageing [26]. Combinations of genetics, genomics, technology, and therapeutic measures can bring improvements to the healthcare system by targeting individual. Precision medicine is another approach which is also targeting the individual. Genetics is the study of heredity. Genomics is the study related to genes and their functions, and techniques related to them. Genetics is about the functioning and composition of a single gene whereas genomics is about all genes and their relationships in order to identify their combined influence on the development of the organism [27]. The coordination and combination of bioinformatics is helpful in developing various tools for the people who are suffering from dementia. Prevention is possible when we know the risk of people to get diseases. Genomic research is becoming widespread by providing information related to risks of getting specific diseases in early stages by studying molecular aspects of the disease. For instance, studying amyloid-beta protein, the apolipoprotein E (APOE) E4 allele (APOE*E4) helps in determining Alzheimer’s disease risk [28].

3.1 Personal genomics

The full genome is the complete DNA sequence containing the complete information of an individual. An allele is one of the several forms of a gene. Different alleles can result in e.g.

different pigmentation. For example, the gene for eye colour has several alleles such as an allele for brown eyes and an allele for blue eyes. Genomics is the study of genes and their functions.

The main aim of genomics is understanding the structure of the genome, including the mapping genes and sequencing the DNA. Personal genomics is a branch of genomics which deals with the analysis of the genome of individuals. There are various new techniques available for identifying genotypes of individuals, like full genome sequencing or single-nucleotide polymorphism (SNP) genotyping and identified genotypes are compared with existing literature in order to predict the risk of genetic disorders. Every genome is unique and new sequencing technologies made available to get the sequenced genomes for individuals with fewer costs.

When the Human Genome Project started it took seven years to sequence the first percent of the human genome, while the second per cent took only one year. Techno-optimists said that genome sequencing becoming cheap enough, it can be used on every person, every cancer cell, every bacterium, and every virus, which helps in predicting the future [29]. Personal genomics is very helpful for identifying genetic inclination of a particular person towards common diseases. If a person or a caregiver or an insurance company knows that one is inclined to a certain disease it will be easier for them to take precautions such as change lifestyle, diet, physical activities and regular health check-ups [30]. The changes in one of these lifestyle- related factors may influence the total risk score.

Personal genomics can be used by health care to educate and counsel the citizens. It can also be used to treat certain diseases with just the right medication, using existing data about how the substance worked on other patients with same genotypes to ensure that the medicine has maximum effect on the disease with no or fewer side effects. Personal genomics can also be used to advise couples who are planning to have children. For example, if both individuals are carriers of a genetic disorder like cystic fibrosis it will increase the probability of the child having

(22)

this disease. This may give a choice to the couple having babies in another way, for example using IVF where embryos are diagnosed in the lab for genetic disorders before implanting in the womb [30].

Mobile applications have been developed to analyse an individual’s DNA by submitting their genome sequence. An example of such an app is GeneG, in which a user can upload their genome sequence to a site and get to know if they have any predispositions for certain diseases by just clicking on a smartphone screen. For example, a pregnant woman needs to donate a sample of DNA for each disease she wants to test since traditional clinics generally process only one gene at a time. This means that the pregnant woman has to go to the clinic every time when she needs to donate sample and takes several days to get the results. Using GeneG user can upload their genome in VCF format (a format used for genome sequence variations) to the site. Once online, they can submit the genetic information to standard analyses developed by organisations like the National Institute of Health, European Bioinformatics, and Standard University. All this can be done remotely by just clicking the app. The information submitted by GeneG user will be kept securely and is under the control of user including who can access the test results. Using information got by GeneG’s testing system doctors can provide successful treatments for various diseases which range from cancer to sleep disorders. At the same time user have more information about their genetic data, they may demand solution according to the results from DNA tests which puts pressure on medical people. Gentle is an iPhone app which offers a DNA-test that includes analysis from 1700 genetic conditions. Many genetic testing companies’ takes the test results and dump into user’s web account without any further counselling. Apart from this Gentle app provides test results information to doctor appointed by the user and further Gentle team provides service for follow-up questioner that may come from user’s doctor. The test data will be transferred securely and user’s data will be kept secretly by only allowing user’s prescribed doctor and geneticists.

The cost of personal genomics is decreasing and dropped to around $1200 in 2015. In Figure 3, the cost curve shows a decrease in genome cost. In 2001 the cost was $95 263 and this was reduced to $1245 by 2015 [31]. Many companies like 23andMe, Navigenics, Pathway genomics, and deCODEme then offered services for genetic testing. Reduction in cost of personal genomics indicates that in near future common people may get their genome sequenced which is helpful in predicting risk related to certain diseases. Risk prediction applications may collect genome information from the user for efficient prediction results.

(23)

Figure 3 Cost per Genome [31]

3.2 Personalised health

Not every disease affects every person in the same way, and even the treatment needs to be varied [32]. In the course of unprecedented scientific innovations and technological enhancements, precision medicine has the ability to detect the disease at its earliest stage (In some cases even before a person is born and still in the stage of the embryo or even before the fertilised egg is implanted on uterus wall) and stop the progression of the disease. The genotyping of drug metabolising enzymes has also helped in improving the drug dosage and thereby reducing the side effects drastically [32 page 4].

Precision medicine uses the molecular markers that signal the symptoms of a disease very earlier then the disease is diagnosed in a clinic. President Obama paid out $215 million in 2015 as an initiative for discovering genetic causes of diseases [33]. As a part of precision medicine initiative President Obama and other officials announced that the National Institutes of Health going to spend $55 million in a single year to find out the genetic and environmental risk factors interaction in causing cancer disease [34]. Many diseases such as dangerous cancers can be

$1 000

$10001 000

$20001 000

$30001 000

$40001 000

$50001 000

$60001 000

$70001 000

$80001 000

$90001 000

$100001 000

Sep-01 Apr-02 Nov-02 Jun-03 Jan-04 Aug-04 Mar-05 Oct-05 May-06 Dec-06 Jul-07 Feb-08 Sep-08 Apr-09 Nov-09 Jun-10 Jan-11 Aug-11 Mar-12 Oct-12 May-13 Dec-13 Jul-14 Feb-15 Sep-15

Cost per Genome 2001- 2015

(24)

totally cured when diagnosed at early stages. Personalised health not only detects the diseases at the very early stage but also shifts the emphasis of medicine from reaction to prevention.

Figure 4 Percentage of patient population where a particular drug in a class is ineffective [32]

Personalised health helps avoid adverse drug reactions. Around 5.3 percent of all hospitalisations are due to drug reactions [35]. Precision medicine has fewer side effects and thereby increased patient adherence to treatment. For example, a study revealed that patients with knowledge of a genetic predisposition for high cholesterol have shown more than 86 percent adherence to their treatment, whereas it was only 36 percent of patients without this knowledge [36]. Precision medicine will also improve the quality of life, for example usually a heart transplant patient needs continuous diagnosis for if immune system is trying to reject the new organ (which can be fatal for the patient) and this diagnosis is very complicated where a pipe is inserted into a vein in neck and threading it towards heart. This entire process simply can be replaced with a simple molecular test which requires a blood sample, making life easier and smoother. Personalised health can help in finding alternative uses for medicines. A medicine which is ineffective on a generalised group of people can be very effective in a specific group. So if a medicine which is ineffective in 90% of cases can still be manufactured for 10% of the population and need not disappear from the production. Precision medicine can help to control the overall costs of health care, for example by avoiding trial and error medicine prescription, by reducing the hospitalisations due to adverse drug reactions and by detecting the diseases very early stage (and saving the cost of drugs).

As every coin has two sides, precision medicine has its own limitations. Some difficulties are expressed by physicians due to uncertainty in genomic test results. There are difficulties in putting precision medicine into clinical practice as participants do not have proper training regarding genomic tests and interpretation of test results. Consideration of only genetic risk factors may not be helpful as dementia can be the effect of both genetic and non-genetic risk factors (ex: age, education etc.). One has to think about risks of sharing personal genomic data before using it.

(25)

Study related to personal genomics, precision medicine is important as particular allele (example APOE*E4) associated with risk of developing dementia. Consideration of personal genomics information in MULTI-MODE app may provide better results which are applicable for individuals.

(26)

4 Risk scores

Various risk scores have been developed to predict the risk of individuals for cognitive decline within a given time frame. These risk scores include only a few well-known risk factors that are easily measurable to calculate the subsequent risk of the event or disease. The main use of these risk scores is targeting of the preventive measures to those who are at risk and are also used to distribute understandable information to the layperson. The developed risk scores have been calculated based on different risk factors. For understanding various risk factors related to dementia, some risk scores have been discussed in this section.

Finish Geriatric Intervention Study (FINGER), CAIDE risk scores were described which are the base for new CAIDE risk score for late-life. Old CIADE risk score app has experimented with selected person’s data and results are shown in section 4.2. ANU-ADRI study has been explained, as study involved with some experiments related to old CAIDE risk score. To explore more about risk factors used in late-life dementia, further desktop research has been conducted and lead to two risk scores LOAD and Late-life dementia risk index, as these are related to late-onset dementia and study related to these is easily available on the internet.

4.1 FINGER

FINGER was a two-year population-based multi-domain randomised controlled trial [37] that tests whether nutritional guidance, mental training, exercise, and reduction of vascular risk factors (for example control of blood sugar levels in people with diabetes) can prevent cognitive impairment and disability.

One cannot change one’s age or family histories but there are some modifiable factors which help in active ageing. A third of Alzheimer’s disease is correlated with some modifiable risk factors such as low education, mid-life hypertension, mid-life obesity, diabetes, physical inactivity, smoking, and depression [38]. Better lifestyle, e.g. due to improved economic status can have a positive impact on dementia in later stages. Physical activity, cognitive training, or both have positive effects on cognition when compared to single-domain prevention trail (testing the efficacy of diet, exercise, cognitive stimulation, and vascular risk factor interventions separately) for cognitive impairment [39].

The FINGER study was the first intervention trial in the world using multi-domain intervention approach for preventing dementia. In this trial, participants aged 60-77 years were selected from previous national surveys. Participants were having, CAIDE (Cardiovascular Risk Factors, Ageing and Dementia) Dementia Risk Score of 6 points or more and cognition at the mean level or less than expected for age. Participants were divided into two groups (1:1 ratio), one group is a multi-domain intervention (diet, exercise, cognitive training, vascular risk monitoring) and another group is control group (regular health advise). A two-year study is done and cognition is measured through neuropsychological test battery (NTB) Z-score. Study is done by providing regular health advice to control group and additional four intervention components related to diet (nutritional intervention), exercise (programs for progressive muscle strength training, aerobic exercise), cognitive training (memory tests, word games), vascular risk monitoring (by checking and advising about blood pressure, weight and BMI (Body Mass Index) to intervention group. Various tests have been done in the two-year study at intervals of 6 months, 12 months and 24 months. After 2-year study results are observed as NTB total score (a measure of drug efficacy used in clinical trials), executive functioning (like planning, problem-solving, attention, decision-making), processing speed, memory are in the much better condition in the

(27)

intervention group compared to control group. So this proof of concept randomised control trial (RCT) showed the importance of multi-domain prevention approaches which also shows beneficial effects of intensity, long duration, type (e.g., multi-domain), and choice of participants (risk individuals) on cognition. FINGER study shows how long multi-domain intervention has its positive effect in preventing or maintaining cognitive decline but does not explain the effect of the individual factor on the overall effect. Considering the effect of individual factor may give beneficial results in developing evidence-based e-health tools which help in promoting healthy brain in ageing.

4.2 CAIDE risk score

The CAIDE risk score was developed to identify the risk level of a person getting dementia in his/her old age. Data for calculating risk score is taken from CAIDE study which is conducted by considering 1409 individuals (875 (62%) were men and 534 (38%) were women from four separate and independent population-based random samples) by examining their mid-life and conducting re-examination 20 years later at late life for later signs of dementia. This study is based on multifactor analysis related to education, age, sex, hypertension, hyperlipidaemia, and obesity [40]. These factors are divided into three categories: vascular (BMI, SBP, DBP, cholesterol), socio-demographic (age, education, sex), and lifestyle characteristics (smoking, physical inactivity). 4% of participants were diagnosed with dementia in those most of the participants were older, less educated, and had vascular risk factors at mid-life compared to the participants who were not diagnosed with dementia [16] .

The first survey conducted by CAIDE included basic questionnaire related to health status, health behaviour and medical history. Serum cholesterol, systolic (the top number of blood pressure) and diastolic blood pressure (the bottom number of blood pressure) was taken. Body- mass index was calculated from measured height and weight. The second survey (Re- examination) was conducted in 1998 after 20 years from the first survey. During the second survey same survey methods adhered from the previous survey and additionally APOE genotypes were analysed. Cognitive status also considered during the second survey.

Risk scores are calculated from β coefficients using logistic regression models, according to the risk factor profiles at middle age. The range of possible risk scores is 0-15 in model 1 (baseline survey) 0-18 in model 2 (2^nd survey). The range is calculated as the sum of a highest possible risk score for each risk factor.

Table 1 CAIDE Risk Score

Risk factor Risk Score

Age <47 years 47-53 years >53 years

0 3 4 Education

>= 10 years 7-9 years 0-6 years

0 2 3 Sex

Female Male

0 1 Systolic Blood Pressure

<= 140 mm Hg 0

(28)

>140 mm Hg 2 Body Mass Index

<= 30 kg/m2 >30 kg/ m2

0 2 Total Cholesterol

<=6.5 mmol/L

>6.5 mmol/L

0 2 Physical Activity

Active Inactive

0 1

Example: Consider a 50-year-old woman who is having 8 years of education, physically active, having systolic blood pressure of >140 mm Hg, body mass index of >30 kg/m2, cholesterol of

>6.5 mmol/L. Risk score for this person is the sum of the highest possible risk score i.e. 11 (3 + 2 + 0 + 2 + 2 + 2 + 0) .

Numerous apps developed using mobile health (mHealth) technologies developing numerous apps for depression, diabetes, cardiovascular diseases and lifestyle changes [41] [42]. To make CAIDE Risk score more accessible to general people and health care practitioners, the researchers who developed CAIDE Risk Score sought to develop a mobile application using mHealth technologies. The result of this is CAIDE Risk Score mobile app which is the first evidence-based App for dementia. CAIDE Risk Score App helps users to detect their individual risk of getting dementia in later 20 years, provides guidance for reducing the risk of getting dementia and also advice for consulting health care practitioner if needed. mHealth apps are helpful in various ways such as keeping track of individual data frequently which helps more data available for health care system for proper medication, tracked data can be used for research purpose in health care systems. One has to consider limitations regarding mHealth apps before releasing into the market. Collected health data from users must be filtered and translated to a message so the user and clinicians understand it. Clinicians and user should be educated before the app being used, to avoid a chance of misunderstanding about the prediction results. Costs involved in developing mHealth tools are also considered as one of the limitation.

CAIDE Risk Score App was released in two versions, one for health care personnel and other for general people [16]. The app is downloaded based on the user selection. Health care practitioner version has an option to enter his/her name, contact details of practice and website information. The app can be downloaded from App Store on iPhone or iPad. Users of App are asked to enter the date of birth, sex, duration of education in years (Figure 5), height, weight, systolic and diastolic blood pressure, serum cholesterol, and physical activity information (Figure 5).

(29)

Figure 5 CAIDE Risk calculator Personal Data.

Based on entered information, App calculates the risk score and displays the risk level in the form of a graph. Graph display will be in various colours based on the risk score. For normal risk score (of 8-9) an orange bar, lower risk score (of 0-9) green bar, higher than average risk score (10-15) a red bar (Figure 6) appears.

Figure 6 CAIDE Risk score calculator level indicator graphs.

One of the tabs in the app has the explanation of risk score and provides probability percentage of developing dementia compare to average probability of developing dementia (Figure 7). App also provides the information on how to modify the risk factors so that individual can reduce the chance of developing dementia (Figure 7).

(30)

Figure 7 CAIDE Risk score calculator information pages.

4.3 Australian National University AD Risk Index (ANU-ADRI) ANU-ADRI Risk Index was developed to identify the degree of older individuals who are at risk of AD and an attempt made to compare the risk index with old CAIDE risk score which was for middle-aged people [17]. This tool differs from previous tools developed to predict dementia, as it was developed by identifying risk factors from three independent older cohorts instead of risk factors from a single cohort study, and did not include variables which require laboratory tests. The tool has wider utility as it was evaluated against both subtypes of dementia like AD and general dementia. An evidence-based medicine approach was selected to identify the risk factors (age, sex, low education, traumatic brain injury, diabetes, low social networks, depressive symptoms, smoking) and protective factors (cognitively stimulating activities, physical activity, alcohol consumption, fish intake) for AD. Developed risk index excluded BMI and. ANU-ADRI risk score had a range of -11 to 56, which involved complete data from 2496 participants. In a study made by ANU-ADRI, CAIDE risk score was calculated and compared in three different older aged cohorts. The comparison of CAIDE risk score did not result in high c- statistics when used on older cohorts. This can be due to CAIDE development was done by considering mid-life risk factors and was validated in older cohorts. The effect of risk factors can be different in different ages. Factors like BMI, and cholesterol less associated with late-life AD risk [43] [44] when compared to mid-life and were excluded from ANU-ADRI.

(31)

4.4 Comparison between old CAIDE, ANU-ADRI and new CAIDE risk scores:

Table 2 Comparison of risk scores

Old CAIDE risk score ANU-ADRI risk index New CAIDE risk score (Developing stage) Based on mid-life risk

factors

Based on late-life risk factors

From single cohort study ( CAIDE)

From three independent cohorts of older adults (MAP, KP, CVHS)

From three independent cohorts of older adults (SNAC-K, CAIDE, KP) Score values were

derived from β coefficients of the logistic regression model.

Score values were derived from β coefficients of the logistic regression model.

Score values were derived from β coefficients of cox regression model.

1409 participants 2496 participants 3744 participants Risk score range is 0

to 15

Risk score range is -11 to 56 Risk score range is 0-15 Preliminary study: Age

39-64 yrs. Follow-up study: 65-80 yrs.

Comparison of adults aged

<70 with >=70 was done

Comparison of adults <74 with >= 75 was planned Strengths:

1. Information gathering done at mid-life and again at late-life 2. Number of

participants was high: more than 80% of the baseline

examination and greater than 70%

at the follow-up examination.

Limitations:

1. The score is applicable to the prediction of dementia risk for the people who survive 20 years after the baseline study. Those who died before follow-up study

Strengths:

1. Data collected from three independent older cohorts.

2. The tool has wider utility as evaluated against subtypes of dementia like AD and general dementia.

3. Risk factors do not need any laboratory tests.

4. Risk index has been compared with pre- developed CAIDE risk score.

Limitations:

1. ANU-ADRI did not validate against younger cohorts.

2. Isolated findings from KP, MAP, and CVHS cohorts were used in some of the meta- analyses but validation

Strengths:

1. Data collected from three independent older cohorts.

2. New risk score has been validated according to TRIPOD rules.

(32)

may had dementia which was not

considered.

2. The family history of dementia, waist-hip ratio and presence of diabetes or insulin resistance, concentrations of high-density and low-density lipoproteins were not considered in calculating risk score which can be a part of developing dementia.

samples were not purely independent.

MAP - Memory and Aging Project, KP - Kungsholmen Project, CVHS - Cardiovascular Health Cognition Study

4.5 Late Onset Alzheimer’s disease risk score

A risk score has been developed for the prediction of Alzheimer’s disease in elderly persons based on vascular risk profiles [45]. Participants for the study were the people aged 65 years or older and were residents of New York who were free of dementia and cognitive decline. Age, sex, education, ethnicity, APOE*E4, diabetes history, high-density lipoprotein levels, hypertension or smoking and waist to hip ratio were considered as risk factors. Several risk factors were explored separately with Late Onset Alzheimer’s disease risk score (LOAD) using Cox proportional hazards models to identify the risk factors which contributes more to the risk score.

All participants in the study underwent an interview related to general health, medical history, and neuropsychological issues. 1051 people participated in the study. For calculating risk score, study chose the similar approach used in CAIDE. A follow-up study conducted after 18 months of base study. Follow-up study resulted in less number of people with dementia compared to the first study. This is due to the demented people from the first study were older, less educated, had a higher prevalence of diabetes, a higher WHR (waist to hip ratio), and lower HDL-C (high-density lipoprotein cholesterol) levels.

Risk score has been categorised into five groups due to large risk score range (0-60). According to developed risk score quintiles, the probability of LOAD was 1.0 persons with a score of 0-14, 3.7 for persons with a score of 15-18, 3.6 for persons with a score of 19-22, 12.6 for persons with a score of 23-28, and 20.5 persons with a score higher than 28. The probability of LOAD was increased with higher vascular risk score and a greater number of risk factors.

(33)

Table 3 LOAD Risk score

Risk factor Risk Score

Sex M F

0 1 Age (yrs.)

65-70 >70-75 >75-80 >80-85 >85

0 6 8 13 21 Diabetes

No Yes

0 3 Hypertension

No Yes

0 1 Current Smoking

No Yes

0 5 Low HDL-C

No Yes

0 3 High WHR

No Yes

0 7 Education (yrs.)

>9 7-9 0-6

0 8 11 Ethnicity

White Black Hispanic

0 5 4 APOE*E4

None >= 1

0 4

This risk score is mainly used to estimate the risk of developing dementia in older people and it can also use in genetic research of dementia to adjust a compound variable of non-genetic risk factors.

4.6 Late-life dementia risk index

The late-life dementia risk index can accurately divide older adults into those with a low, moderate, or high risk of developing dementia within 6 years [46]. 3,375 participants were involved in the Cardiovascular Health Cognition study [47] without evidence of dementia at baseline. Logistic regression models are used to identify the risk factors which are most predictive of developing dementia within 6 years and developed a point base system. The

(34)

point’s possible range is 0-15. Participants mean age baseline was 76 years: 59% were woman and 15% were African American. Fourteen percent (n=480) of the participants developed dementia within 6 years. The developed point based system includes the risk factors age, poor cognitive test performance, body mass index, APOE*E4, cerebral MRI findings of white matter disease, internal carotid artery thickening on ultrasound, slow physical performance, history of bypass surgery, lack of alcohol consumption. 4% of the participants subjected with low scores, 23% of participants subjected with moderate scores and 56% of participants subjected with high scores.

Table 4 Late-life dementia risk score

Characteristic Points

Age

75-79 (yrs.) 80-100 (yrs.)

1 2

Low 3MS 2

Low DSST 2

BMI < 18.5 2

>=1 APOE*E4 1

MRI white matter disease (grade>=3 1 MRI enlarged ventricles (grade >=4) 1 Internal carotid artery thickening

>=2.2mm

1 History of coronary bypass surgery 1 Time to put on and button shirt >45s 1 Lack of alcohol consumption 1 Age is compared to those who aged 65 to 74 years.

Low 3MS: <= 87 (all white participants and black participants with >=12 years education) or

<=70 (black participants with <12 years of education).

Low DSST: <=33 (white participants with >=12 years of education) or <=22 (white participants with <12 years of education and all black participants).

3MS: Modified Mini-Mental State Examination; DSST: Digit Symbol Substitution; BMI: Body Mass Index; CI: Confidence interval.

The late-life dementia risk index tool can be used in clinical or research area to target prevention and intervention strategies towards high-risk individuals. The late-life dementia risk index also used to identify the older adults who should be monitored for the new dementia symptoms, so that treatments could be initiated at the earliest possible stage of the diseases, and helps in providing information to concerned patients or family members. The late-life dementia risk index could be used to reassure the older adults who do not currently have overt dementia and to provide those individuals whose risk is high with information that may helpful in planning for better future.

4.7 MULTI-MODE

To promote healthy living, improve health care, and support active ageing, EIT (a part of the European Union Horizon 2020 effort) has launched EIT Health. EIT Health integrates higher education, research, and business for the enhancement of healthy living and to improve health

Using non-medical risk factors related to dementia and cognitive decline for developing an evidencebased e-health tool

Using non-medical risk factors related to dementia and cognitive decline for developing an

evidencebased e-health tool

ANUSHARANI GOPU

Using non-medical risk factors related to dementia and cognitive decline for developing an evidence-based

e-health tool

Anusharani Gopu

Supervisor: Marie Sjölinder

Examiner: Magnus Boman

Acknowledgement

Abstract

Keywords

Sammanfattning

Nyckelord

Table of contents

List of figures

List of tables

List of abbreviations

1 Introduction

2 The role of ICT in active ageing and risk factor categorisation

3 Targeting the individual

Cost per Genome 2001- 2015

4 Risk scores